Modelpublic

hi-space/GR00T-N1.6-3B-Pick-Orange

sourceHugging Faceotherupdated 5mo agoView on Hugging Face

1likes19downloads

Model Card

GR00T-N1.6-3B-Pick-Orange-Sim

A fine-tuned version of nvidia/GR00T-N1.5-3B for orange pick-and-place tasks in simulation, trained on the LightwheelAI/leisaac-pick-orange dataset.

Model Description

GR00T-N1.6 (Gr00tN1d6) is a vision-language-action (VLA) model for robot manipulation. This checkpoint is fine-tuned for a pick-and-place task where the robot picks up an orange in a simulated environment.

—Architecture: Gr00tN1d6 with Eagle-Block2A-2B-v2 vision-language backbone + diffusion policy action head
—Base model: nvidia/GR00T-N1.5-3B
—Task: Pick orange (simulation)
—Action horizon: 50 steps
—Inference timesteps: 4 (diffusion)
—Model dtype: bfloat16

Fine-tuning Configuration

Parameter	Value
Tuned components	Diffusion model, projector, top 4 LLM layers, VL-LN
Frozen components	Vision encoder, LLM backbone
Training steps	6000
Batch size	32
Final training loss	~0.017
Action representation	Relative actions
Attention	Flash Attention 2

Training Details

—Dataset: LightwheelAI/leisaac-pick-orange
—Max steps: 6000 (1 epoch)
—Loss curve: Started at ~1.1, converged to ~0.017

Usage

python

from gr00t.model.gr00t_n1 import GR00TPolicy

policy = GR00TPolicy.from_pretrained("hi-space/GR00T-N1.6-3B-Pick-Orange-Sim")

Refer to the NVIDIA Isaac GR00T repository for full inference and deployment instructions.

Intended Use

This model is intended for simulation-based robotic pick-and-place tasks involving oranges. It is not guaranteed to transfer zero-shot to real hardware without additional fine-tuning.

License

This model inherits the license from the base model nvidia/GR00T-N1.5-3B. Please refer to NVIDIA's terms for usage restrictions.